Meta · Published 2025-06-19 · 2 min read
How I tested Facebook ads: concepts
In April I ran a test campaign to show how Facebook advertising actually works. The details were in the April post. Tomorrow I will show the results a month on; today it is the approach itself.
Working with creatives. I am not a fan of dumping 100+ creatives into the account and hoping Facebook finds the winner on its own. It can work in the short term, but:
How do you scale the success? The chance that creative number 101 beats all the others is exactly 1%. That is plain probability.
Question: what is the probability that out of 100 creatives this particular one turns out to be the best? Answer: one in a hundred.
With that approach many teams lose the incentive to make new creatives once they have already shipped a hundred. Once they have shipped five hundred, there is no incentive at all. Not everyone phrases the problem in terms of probability, but subconsciously everyone understands the arithmetic.
The alternative is thinking in concepts
What is a concept? An example for a video editing app — and not even a made-up one, these are the concepts I would genuinely start with.
- TikTokers
Audience: under 30. What we sell: show yourself dancing. Creative: a 16-year-old dancer, face swapped for yours.
- Parents
Audience: people photographing their kids. What we sell: the Miyazaki effect, a cartoon video with your child in it.
- Nerds
Audience: technical people who care about features. What we sell: trim/upscale plus export to Google Drive.
Two creatives per concept. That is enough for a test.
Expectations.
You can already price the traffic while you are still writing the concepts down
- TikTokers
installs will be cheap, purchase conversion low.
- Nerds
installs will be expensive, purchase conversion high.
- Parents
hard to say. Have to look.
Which audience the economics close on is what the experiment is for.
I do not like mixing concepts inside one ad set. For me six creatives built on an understanding of the audience are a stronger hypothesis than «let us knock out a hundred creatives, something will fire». If one ad set tests different concepts, the data averages out and stops meaning anything. And I need the economics per audience, so I know roughly where to put the money.
📊 Results
I started with about 15 creatives across 3 concepts. As expected
- 90% of the budget went to the top 2 variants
- one of them I picked as the main concept
- and put all the traffic behind it
🔥 Tomorrow — numbers only.